Register-currency preflight: surface every stale register in one pass, not one per release cycle

Finding from the Build J re-cert night (2026-07-10): four gate classes failed SERIALLY across four ~90-minute release passes — #334 Medicaid (RI source hole), #337 parity staleness (5 rotted exemptions), zero-support (5 newly-materialized TANF cells), #369 coverage staleness (same 5 columns) — all one root cause: the week's coverage growth outran the registers, and each gate only surfaces at its pipeline stage, so discovery cost 4 full passes. Proposal: a register-currency preflight — run EVERY register's anti-rot check against the candidate frame in one cheap pass (frame-level, pre-materialization where possible; the #384 cross-register consistency preflight is the pattern, extended to gate-vs-frame currency: parity known-gaps, degenerate exclusions, coverage manifest, zero-support list, take-up contract). One run, one combined report of every stale register, instead of serial ~90-minute discoveries. Refs: #368 (the night's cycle), #384, #369, #337, #386/#387.

Read the cross-register consistency preflight pattern in #384, then review the parity, degenerate-exclusion, coverage, zero-support, and take-up registers named here. Done means one candidate-frame preflight produces a combined report covering every stale register instead of discovering them across separate release passes.